Thursday, May 14 on the Edge of AI. OpenAI is reportedly preparing to sue Apple over a failed partnership, Cerebas just pulled off the year's biggest IPO, and Codex is finally coming to your phone. Let's get into it.
OpenAI, the lab behind ChatGPT, is reportedly preparing legal action against Apple over their AI partnership [6][12]. Bloomberg says the company has hired an outside law firm to work through its options. The core claim: Apple hasn't invested enough in the deal, leaving OpenAI feeling burned after handing over prime access to its models. This isn't the first partner to walk away feeling that way — OpenAI's relationship with Microsoft has been rocky for months, and now Apple is in the same position. The signal: the partnership model for frontier AI is fraying. Exclusive access to a frontier model no longer guarantees the partner will pour resources into it, and when they don't, the lab sues. My read: this is a sign that OpenAI's revenue pressure is real enough that it can't afford a passive partner. If Apple fights or settles, the terms of every lab-platform deal just got a lot harder to negotiate.
Pivoting. Cerebras, the AI chip startup, raised $5.5 billion in the year's largest IPO and saw its stock pop 68 percent on debut [10][13]. Shares closed at $311, well above the $185 IPO price, after being halted for volatility. The offering priced above a revised range and raised nearly 60 percent more than its target. Early venture investors Benchmark and Eclipse are looking at billions in returns — Benchmark's stake alone is worth about $3.2 billion, a twelve-fold return on its $268 million investment [4]. CEO Andrew Feldman, a serial entrepreneur who literally grew up on Stanford's campus, is now worth $3.2 billion himself [11]. What this changes: the AI chip IPO window is officially open. Cerebras proved there's appetite for pure-play AI silicon, not just the big GPU makers. Every startup in the inference or training-chip space just got a benchmark for their own exit.
Now, on the product side. OpenAI is putting Codex, its desktop coding agent, into the ChatGPT mobile app [1]. The move follows the surge in popularity of Anthropic's Claude Code, and OpenAI has been cutting back on "side quests" — shutting down projects like Sora — to focus on enterprise growth. Codex on mobile means you can run app operations from your phone, a step toward the always-on agent vision. The angle: the agent wars are now a mobile battlefield. Whoever gets their coding agent into the user's pocket first wins the habit.
Last beat. Microsoft is planning to remove most of its Claude Code licenses and push developers back to its own tools [3]. The company first opened access to Claude Code in December, inviting thousands of employees to use Anthropic's coding tool. It proved popular — maybe too popular. Microsoft is walking it back, which means the internal clash between using a rival's AI and building your own is now playing out inside the largest software company on earth. The signal: platform incumbents are realizing they can't afford to let their own developers get hooked on a competitor's agent.
Three stories, three different kinds of pressure. Legal, financial, and platform. The common thread: every lab is fighting for position before the window closes.
That is the edge for today.